Selection of cyclic peptide aptamers to HCV IRES RNA
using mRNA display
Alexander Litovchick and Jack W. Szostak*
Howard Hughes Medical Institute, Center for Computational and Integrative Biology, and Department of Molecular Biology, Simches Research Center,
Massachusetts General Hospital, 185 Cambridge Street, Boston, MA 02114
The hepatitis C virus (HCV) is a positive strand RNA flavivirus that
is a major causative agent of serious liver disease, making new
treatment modalities an urgent priority. Because HCV translation
initiation occurs by a mechanism that is fundamentally distinct
from that of host mRNAs, it is an attractive target for drug
discovery. The translation of HCV mRNA is initiated from an
internal ribosomal entry site (IRES), independent of cap and poly(A)
recognition and bypassing eIF4F complex formation. We used
mRNA display selection technology combined with a simple and
robust cyclization procedure to screen a peptide library of >1013
different sequences and isolate cyclic peptides that bind with high
affinity and specificity to HCV IRES RNA. The best peptide binds the
IRES with subnanomolar affinity, and a specificity of at least
100-fold relative to binding to several other RNAs of similar length.
The peptide specifically inhibits HCV IRES-initiated translation in
vitro with no detectable effect on normal cap-dependent translation initiation. An 8-aa cyclic peptide retains most of the activity of
the full-length 27-aa bicyclic peptide. These peptides may be useful
tools for the study of HCV translation and may have potential for
further development as an anti-HCV drug.
hepatitis C 兩 internal ribosomal entry site 兩 RNA binding 兩
translation initiation 兩 viral mRNA translation
H
epatitis C virus (HCV) is a persistent flavivirus that infects
⬇3% of the human population, making it the most common
chronic blood-borne infection (1). Among six genotypes of the
virus, genotype 1 is the most common in Europe and North
America (2). Approximately 75% of HCV-infected individuals
develop a largely asymptomatic chronic infection, whereas
⬇25% of patients eventually develop liver cirrhosis or hepatocellular carcinoma (1). At present, HCV infection is treated by
IFN ␣/ribavirin therapy until sustained viral response (SVR) is
reached. This treatment is effective in ⬇50% of patients (1, 3).
Resistance to IFN and ribavirin (4), and HCV persistence after
SVR in the form of ‘‘occult’’ hepatitis C (3, 5) makes HCV
frequently incurable. The virus effectively avoids the host immune response and no vaccine for hepatitis C is currently
available (6, 7). Significant effort is being put into the development of specifically targeted antiviral therapies for HCV treatment (STAT-C), aimed at different stages of the viral life cycle,
including inhibitors of NS3/4 protease, NS5B replicase, etc., as
well as discovery of an anti-HCV vaccine based on E1/E2 fusion
proteins (7). Emerging drug resistance has already been observed with anti-HCV compounds in clinical trials (4, 7), highlighting the need for new approaches to HCV therapies.
The 340-nt 5⬘ UTR is among the most conserved parts of the
HCV genome. It contains a highly structured internal ribosomal
entry site (IRES) (supporting information (SI) Fig. S1 A) that
mediates the initiation of translation of the viral polyprotein by
a cap- and poly(A)-independent mechanism (8, 9). Translation
initiation from the HCV IRES does not require the eIF4F
complex: the IRES is recognized directly by the 40S ribosomal
subunit and eIF3, recruits eIF2/GTP/Met-tRNA, and the resulting 48S complex assembles at the initiation codon (9, 10). It is
noteworthy that the pathway of IFN inhibition of viral replica-
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0805837105
tion occurs via an IRES-dependent mechanism (11). Both the
IRES structure and the mechanism of HCV translation initiation (e.g., 12, 13) have been the subject of intense research in
recent years as a therapeutic target (14–17). As an example, the
synthetic steroid mifepristone specifically inhibits in vitro translation initiation from the HCV IRES (18). Unfortunately,
mifepristone did not meet the efficacy endpoint for treating
HCV infection in a Phase II clinical trial (19).
We have attempted to isolate more effective and specific
inhibitors of HCV IRES-initiated translation. Here, we describe
the selection of high-affinity peptide binders to the HCV IRES
from a cyclic peptide-mRNA fusion (20, 21) library of 10 trillion
individual sequences. After 11 rounds of selection, we isolated a
bicyclic peptide that binds the HCV IRES tightly and specifically, and selectively inhibits the IRES-initiated translation of a
reporter gene in vitro.
Results
In Vitro Selection. The scheme that we used to select for cyclic
peptides that bind specifically to the HCV IRES is outlined in
Fig. 1. We started the selection with ⬇17 ␮g of a double stranded
DNA library (2.5 ⫻ 1014 individual sequences) designed to code
for his-tagged peptides containing 10 random residues flanked
by cysteines (Fig. S1B). After the transcription of mRNA,
photo-cross-linking to the peptide-accepting 3⬘-puromycin oligonucleotide, and purification by denaturing PAGE, ⬇7 ␮g of
the mRNA-oligonucleotide-puromycin library was obtained
(2 ⫻ 1014 individual sequences). After translation, purification,
and peptide cyclization, ⬇6.5% of the library was converted into
mRNA fusions to cyclic peptides, for an initial library complexity
of ⬇1.3 ⫻ 1013 sequences.
The selection procedure was divided into three phases of
gradually increasing stringency (Table S1; SI Appendix). Rounds
1 and 2 were designed to decrease the complexity of the starting
library while retaining all possible IRES binders. During these
rounds Torula yeast RNA (TYR) was used as a binding competitor, and all column matrix bound material was eluted nonspecifically with NaOH and 8 M urea. Rounds 3–7 were designed
to select more specifically for IRES binders by eluting column
bound mRNA-peptide fusions competitively with soluble IRES.
Following seven rounds of selection, ⬇50% of the input library
was eluted from IRES column with soluble IRES RNA in 2 h
(Fig. S2). The output cDNA of rounds 6 and 7 was cloned and
sequenced, revealing the high complexity of the remaining
library. We then further increased the stringency of the selection
Author contributions: A.L. and J.W.S. designed research; A.L. performed research; A.L. and
J.W.S. analyzed data; and A.L. and J.W.S. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
*To whom correspondence should be addressed. E-mail: [email protected].
edu.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0805837105/DCSupplemental.
© 2008 by The National Academy of Sciences of the USA
PNAS 兩 October 7, 2008 兩 vol. 105 兩 no. 40 兩 15293–15298
BIOCHEMISTRY
Edited by Jennifer A. Doudna, University of California, Berkeley, CA, and approved August 12, 2008 (received for review June 17, 2008)
Fig. 1. Overview of the mRNA display selection procedure. mRNA is conjugated to an oligonucleotide with a 3⬘-puromycin residue; in vitro translation
results in peptide synthesis and covalent transfer of the peptide to the
puromycin. The mRNA displayed peptide is then cyclized by reaction with
dibromo-m-xylene, and the mRNA protected by reverse transcription. Cyclic
peptides that bind to the HCV IRES are selected, and the attached cDNA
amplified and PCR and T7 transcription.
by using additional selection steps in rounds 8–11, aimed at the
selection of peptides with slower dissociation rates, and improved IRES selectivity. Cloning and sequencing of the library
after round 11 revealed that the sequence 6B4, encoding the
peptide MKCSRGIRCAGVLCGSVGHHHHHHHRL, accounted for ⬎30% of the clones. The same sequence was
observed in only 2 of 105 sequences from rounds 6 and 7.
Structure of Peptide 6B4. Inspection of the peptide 6B4 sequence
reveals 3 cysteines, of which 2 originate from the library design,
and an additional one from the random region. This suggests the
possibility of double cyclization of the peptide in the reaction
with dibromoxylene, an additional acceptor of the alkylation
reaction being one of the histidine moieties (M. C. Hartman,
personal communication; ref. 28). To test this hypothesis, the
6B4 peptide was translated in vitro by using the PURE system
(23) and the dibromoxylene cyclization reaction was performed
during peptide purification on a Ni-NTA column (24). MALDI
TOF analysis revealed a molecular weight of 3,054 for the linear
peptide (formylated) and 3,258 for the cyclic peptide (6B4C),
corresponding to the addition of two xylene moieties to the
peptide (Fig. 2A). When the 6B4 peptide was synthesized by
F-moc chemistry, it exhibited a similar double-cyclization product after treatment with dibromoxylene on a Ni-NTA resin (Fig.
S3). The synthetic 6B4 was not formylated at the N terminus and
the molecular weight of the peptide, which is 3,026 for linear
form and 3,230 for the cyclic form, is therefore 28 lower than that
of the corresponding in vitro translated peptide (Fig. S3).
Configuration of Cyclic Peptide 6B4C. Because peptide 6B4 contains
three cysteine moieties, multiple double-cyclization variants of
the peptide are possible, assuming histidine reactivity (Fig. 2B).
Examining the sequence of 6B4 we found a unique pepsin
cleavage site, GVL, between the second and third cysteines. Only
one bicyclic configuration, denoted C1-C2, C3-H (Fig. 2B)
would produce two separate cyclic peptides with molecular
weights of 1,495.7 and 1,753.8 after pepsin cleavage at pH ⱕ 2;
no other configuration would generate two separate products.
MALDI-TOF analysis confirmed that the mass peak of 3,230
completely disappeared from the spectrum, whereas new peaks
of 1,495.4 and 1,753.5 appeared (Fig. 2C). We conclude that the
major configuration of the peptide is C1-C2, C3-H, although
other variants may be present at substantially lower quantities in
the peptide mixture.
15294 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0805837105
Fig. 2. MALDI-TOF analysis of 6B4 peptide. (A) Mass spectra of in vitro
translated peptides. Red trace: linear 27-mer 6B4, 3,053.85 observed, 3,054.5
expected. Blue trace, cyclic 6B4C, 3,258.6 observed, 3,258.5 expected. The mass
peak at 3,495 (⫹ve ion mode) is the insulin B chain calibration standard. (B)
Possible configurations of bicyclic 6B4C peptide, assuming that one cyclization
event involves two cysteines, and the second cyclization involves the remaining cysteine and one histidine. Only the C1-C2, C3-H configuration will be
cleaved into two separate cyclic peptides by pepsin digestion. Predicted
masses are shown. (C) MALDI-TOF mass spectra of synthetic, cyclized 6B4C
peptide (top red trace, observed 3,230.48, expected 3,230.5), pepsin-digested
6B4C (blue trace, note absence of uncleaved peptide at mass peak 3,230.5, and
presence of predicted C1-C2 fragment (observed 1,495.4, expected 1,495.7)
and C3-H fragment (observed 1,753.5, predicted 1,753.8). The greater intensity of the peak corresponding to the N-terminal C1-C2 fragment may reflect
enhanced ionization due to the presence of cationic residues in the N-terminal
peptide loop. The lower pink trace is the spectrum of a pepsin self-digest,
showing that some of the peaks in the peptide digest originate from pepsin
(external mass calibration standard with insulin B chain mass peak, 3,495; data
not shown).
Binding to HCV IRES RNA. Radiolabeled peptide in both cyclic and
linear forms was used to measure binding affinity to HCV IRES
RNA and other nonspecific RNAs by equilibrium ultrafiltration
(Fig. 3A and Table 1; see also Figs. S4 and S5). The Kd of the
linear peptide for HCV IRES RNA was 6.5 nM, whereas the Kd
of the cyclic peptide was 0.70 nM. Affinities of the cyclic peptide
6B4C to nonspecific RNA targets such as TYR and a 319-nt-long
mRNA (CW mRNA) were 140 and 98 nM, respectively.
We then prepared a linear fluorescein-labeled peptide Fl-6B4
for use in competitive binding experiments to allow for the
measurement of affinities of unlabeled synthetic peptide variants
of the selected peptide. Equilibrium ultrafiltration experiments
by using Fl-6B4 and varying concentrations of HCV IRE RNA
revealed a Kd of 3.0 nM. The Kd values obtained by competitive
binding assays for chemically synthesized 6B4 in linear and cyclic
Litovchick and Szostak
Fraction bound
0.5
0.4
0.3
0.2
0.1
0
0.001 0.01
0.1
1
10
100
1000
nM IRES
1
B
Fraction bound
0.8
0.6
0.4
0.2
0
0.001 0.01
0.1
1
10
100 1000 10
4
pe ptide s, nM
Fig. 3. Binding of peptides to HCV IRES RNA. Binding was measured by
equilibrium ultrafiltration, and plotted as fraction bound vs. concentration of
IRES or competitor peptides. (A) Direct binding of 35S-labeled linear (triangles,
interrupted line) and cyclic (circles, solid line) 6B4 peptide to HCV IRES RNA,
yielding observed Kd of 6.5 ⫾ 1.8 nM (linear peptide) and 0.70 ⫾ 0.14 nM (cyclic
peptide). (B) Competitive binding of linear (triangles, interrupted line) and
cyclic (circles, solid line) synthetic 6B48 peptide to IRES, measured as competition with the fluorescent Fl-6B4 peptide. Observed IC50s are 32 nM for cyclic
6B48C and 102 nM for linear 6B48. Kd as calculated from competition with
0.625 nM Fl-6B4 peptide and 15 nM IRES RNA (see Materials and Methods and
Fig. S7) are 3.3 ⫾ 0.8 nM (linear peptide) and 0.65 ⫾ 0.12 nM (cyclic peptide).
Values are the mean and standard deviation of three to five Kd measurements
in each experiment.
form were 3.3 nM and 0.65 nM (Table 1), respectively, similar to
the values obtained by the direct binding assay by using radiolabeled 6B4 and 6B4C. We also measured the affinities of 6B4
and 6B4C for RRE and class I ligase RNAs, which were not used
during the selection (Table 1). The linear 6B4 peptide exhibited
only moderate specificity (8- and 70-fold tighter IRES binding vs.
RRE and class I ligase, respectively), whereas the cyclic 6B4C
peptide was highly specific (⬇150- and 300-fold, respectively).
Inhibition of IRES-Mediated Translation Initiation In Vitro. To determine whether the selected peptide can specifically inhibit translation initiated from the HCV IRES, we prepared two constructs
with Gaussia Luciferase (GLuc) as a reporter gene. In one
construct translation of the GLuc gene was placed under the
Minimal Active Structure of 6B4 Peptide. To identify the minimal
region of the selected 6B4 peptide responsible for the HCV
IRES binding and translation inhibition properties, we performed experiments with shorter synthetic versions of the 6B4
peptide. Based on the results of pepsin digestion of the 6B4C, we
synthesized the N-terminal loop of this peptide, an 8-mer
KCSRGIRC (referred to as 6B48), and performed the cyclization reaction in solution (to generate 6B48C). We studied the
affinity of the purified linear and cyclic peptides for HCV IRES
RNA and their potential as inhibitors of IRES-initiated translation (Table 1). The 8-mer peptide was 3- to 5-fold weaker in
affinity for the IRES than the full-length 27-mer 6B4 or 6B4C,
with a Kd of 17.5 nM in linear and 3.7 nM in cyclic form (Fig. 3B,
Table 1). The specificity of this shorter-cyclic peptide is similar
to that of full-length 6B4C. Kd measured by competitive binding
for 6B48C to a 319-mer mRNA and ribosomal RNA were found
to be 161 and 276 nM, respectively. Affinities to RRE RNA and
class I ligase were ⬇120 nM and 0.8 ␮M, respectively (Table 1).
Both linear 6B48 and cyclic 6B48C peptides specifically inhibit
IRES-initiated translation in HeLa extracts (Fig. 4 D and E). The
IC50 of linear 6B48 for IRES-initiated translation inhibition was
found to be 125 nM, and for 6B48C it was 76 nM, at an mRNA
concentration of 50 nM (Table 1, Fig. 4D). The translation of
leader-GLuc and capped leader-GLuc constructs was not inhibited in HeLa extracts by up to 2 ␮M these peptides (Fig. 4E). The
similar IC50 values observed for all peptides in translation
inhibition assays reflect the fact that the IRES-mRNA construct
used in the translation reactions was present at a concentration
well above the Kd for the peptides.
Discussion
The isolation of highly specific peptide aptamers directed to
pharmacologically relevant RNA targets has been an ongoing
challenge. Although high-affinity cationic linear peptides are
Table 1. Structure–activity relationship of 6B4 variants
Compound
6B4
6B4C
6B48
6B48C
Mifepristone
Mr
IRES Kd, nM
3,026.4 3,053.9*
3,230.5 3,258.6*
923.4
1,024.99
429.6
(6.5 ⫾
3.3 ⫾
0.65 ⫾ 0.12† (0.70 ⫾ 0.14‡)
17.5 ⫾ 2.7†
3.7 ⫾ 0.5†
N/A
0.8†
1.8‡)
RRE Kd, nM
Class 1 ligase
Kd, nM
IC50 HeLa, nM
27 ⫾ 4
114 ⫾ 5
206 ⫾ 1
118 ⫾ 9
N/A
219 ⫾ 20
228 ⫾ 24
818 ⫾ 81
1,109 ⫾ 109
N/A
95
64
125
76
1,200
*Formylated peptide synthesized by in vitro translation.
†Measured by competitive binding by using fluorescent Fl-6B4 peptide.
‡Measured by direct binding using radiolabeled 6B4 peptide.
Litovchick and Szostak
PNAS 兩 October 7, 2008 兩 vol. 105 兩 no. 40 兩 15295
BIOCHEMISTRY
control of the HCV IRES and, in the other construct, translation
was controlled by a generic consensus leader sequence carrying
a 5⬘ m7GpppG cap analog (Fig. 4A). In HeLa cell translation
extract the IRES-GLuc and the m7GpppG cap analog leaderGLuc mRNA constructs exhibited similar levels of translation,
whereas the uncapped mRNA construct was translated very
inefficiently (Fig. S6A). The linear and cyclic peptides 6B4 and
6B4C were found to selectively inhibit IRES-initiated translation
of GLuc with IC50s of 95 and 64 nM, respectively, at an mRNA
concentration of 50 nM (Fig. 4B, Table 1). Translation of capped
leader-GLuc was not inhibited by up to 5 ␮M peptide (Fig. 4C).
Thus, the inhibition of translation by 6B4C in vitro was specific
for IRES-initiated translation. As a positive control for both
experiments we used 1–2 ␮M mifepristone, which has previously
been shown to inhibit HCV IRES-mediated translation (18).
Mifepristone produced ⬇50% inhibition of Gaussia Luciferase
translation from the IRES-GLuc construct while not affecting
the translation of the leader-GLuc constructs.
0.6
A
A
m7Gppp
GLuc
Leader
B 100%
C
% translation
6B4C
6B4
80%
% translation
GGTCAGATCCGCTAGCGCTACCGGTCGCCACC
GLuc
HCV IRES
60%
40%
80%
6B4C
6B4
60%
40%
0%
0%
0
20
50
100
200
peptide, nM
100%
400
1000
5000
peptide nM
E
6B48C
6B48
80%
0
500 1000
120%
% translation
% translation
100%
20%
20%
D
120%
60%
40%
20%
100%
80%
6B48C
6B48
60%
40%
20%
0%
0%
0
20
50
100
200
500
1000
peptide, nM
0
400
2000
peptide, nM
Fig. 4. Selective inhibition of IRES driven translation by linear and cyclic 6B4 and 6B48 peptides. (A) Schematics of Gaussia Luciferase reporter constructs. (Left)
HCV IRES-Gluc construct, (Right) control capped Gluc mRNA. (B and C) Translation of IRES-GLuc (B) and capped leader-GLuc (C) in HeLa S10 extract in the presence
of linear 6B4 and cyclic 6B4C, plotted as percent of GLuc luminescence relative to the untreated control. (D and E) Translation of IRES-GLuc (D) and capped
leader-GLuc (E) in HeLa S10 extract. in the presence of linear 8-mer (6B48) and the cyclized 8-mer (6B48C) plotted as percent of GLuc luminescence relative to
the untreated control.
easily obtained, they are often not very specific for the particular
target RNA sequence, because the electrostatic interaction of
the cationic side chains with the anionic RNA backbone typically
dominates the binding (26). Recent work on cyclic peptides
suggests that rigid structures may yield selective binding not
dominated by electrostatics (e.g., 29, 30). We have attempted to
address the specificity problem by selection from a very large
library of cyclic peptide-mRNA fusions under conditions that
stringently select against nonspecific binding. We used cyclized
peptides to minimize the entropic cost of peptide binding to the
target RNA. To ensure high affinity and specificity of the
selected peptide aptamers, the binding selection was carried out
in the presence of high concentrations of salt and arginine to
reduce nonspecific electrostatic interactions, along with high
concentrations of competitor RNA to minimize nonspecific
binding. We also used competitive elution of the fusions from the
selection column with soluble IRES RNA as a further selection
for specific binding. After seven rounds of the selection, we
observed a large number of peptide sequences (including 6B4)
allowing for high-affinity IRES RNA binding (27), and after four
more rounds of selection the highly specific peptide 6B4 accounted for 30% of the surviving sequences.
The selected 6B4 peptide has several noteworthy features.
Most striking is the presence of a cysteine residue at a position
in the peptide derived from the random region of the original
peptide library. This additional cysteine moiety allows for double
cyclization after reaction with dibromoxylene: two of the cysteines form one loop, whereas the third cysteine and a histidine
(28) form the second. A unique pepsin cleavage site in the 6B4
sequence allowed for unambiguous assignment of the structure
of the bicyclic 6B4C: the first loop is between the first two
cysteines, and the second loop is between the third cysteine and
one of the histidine residues of the his-tag. The factors that drive
cyclization into this particular structure are unknown, but may
include preorganization of the peptide structure, or greater
steric accessibility of the N-terminal region of the peptide when
15296 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0805837105
the peptide is immobilized on a Ni-NTA resin via its C-terminal
his-tag. A chemically synthesized N-terminal 8-mer peptide
bound to IRES RNA almost as well as the full-length 27-mer
peptide, suggesting that the N-terminal region contains essentially all of the specificity determinants of the selected peptide.
The N-terminal 8-mer peptide contains three basic residues,
which are likely to contribute to binding by interaction with
specific phosphates in the folded RNA structure. The approximately threefold weaker IRES binding of the 8-mer vs. the
original 27-mer may reflect the loss of interactions with the
his-tag portion of the peptide. The approximately fivefold tighter
IRES binding of the cyclic compared with the linear peptides
probably reflects an entropic binding advantage for the conformationally constrained cyclic peptides. The cyclic peptides also
exhibited much greater IRES specificity than the corresponding
linear peptides, based on a comparison of binding to IRES RNA
and to two highly structured RNA molecules that were not used
in the selection (RRE RNA and class I ligase ribozyme RNA)
(Table 1). These observations reinforce the need for cyclization
to obtain highly specific RNA-binding peptides.
The cyclic 6B4C peptide was selected solely on the basis of
high affinity and specificity binding to the IRES RNA; it is
therefore quite striking that binding does indeed lead to the
specific inhibition of IRES-mediated translation initiation. This
suggests that the IRES RNA contains a highly structured region
that is essential for function, and that acts as an epitope that is
particularly suitable for binding to a structured ligand. Peptide
binding could inhibit translation initiation by simple steric
blockade of interaction with the translational apparatus. Alternatively, ligand binding could prevent an essential conformational transition of the IRES RNA. Future experiments will
address the site of peptide binding to IRES and the mechanism
of inhibition of IRES-mediated translation initiation.
Our results demonstrate that high-affinity, high-specificity
peptide aptamers can be isolated from a sufficiently large
starting library by in vitro selection as long as stringent and
Litovchick and Szostak
Materials and Methods
HCV IRES RNA. Cloned HCV IRES of genotype 1b (nucleotides 40 –372) was a gift
from J. Doudna (Berkeley). The 5⬘-terminal 40 nucleotides (a stem-loop) were
added by PCR to ensure cotranscriptional folding of IRES RNA. A minimum of
1 mM Mg2⫹ was maintained in all IRES dilution buffers to stabilize the folded
state of the RNA. Twenty-three nucleotides were added by PCR on the 3⬘ end
of the construct to ensure synthesis of the fully functional IRES (Fig. S1 A). IRES
RNA (nucleotides 1–395), including the complete HCV 5⬘ UTR and 54 nucleotides of the HCV coding region, was prepared by in vitro transcription by using
T7 RNA polymerase. The immobilized IRES RNA selection column was generated by transcription of IRES 1–395 with GTP-␥-S followed by covalent attachment to iodoacetyl-activated cross-linked acrylamide resin (Pierce); the resin
was then quenched by reaction with mercaptoethanol. The concentration of
IRES immobilized on the column was estimated by immobilization of radiolabeled IRES to be 7.5–10 nmol/ml of resin. HIV RRE RNA and the class I ligase
RNA are described in the SI Text.
Selection Library Synthesis. The DNA library included a random 30-nt region
flanked by cysteine codons (see Fig. S1B for details) and was synthesized by the
Keck facility at Yale University. Transcription, in vitro translation, and mRNApeptide fusion formation were done essentially as described in ref. 21 with minor
modifications as follows. For round 1, 2 nmol of cross-linked mRNA was translated
in vitro in 4 ml of wheat germ (WG) extract (Promega), instead of 10 ml of RRL,
for 1 h at 30°C. Cyclization was performed on an oligo(dT) cellulose (NEB) column
equilibrated with cyclization buffer [660 mM KCl, 20 mM Tris, pH 8.0, 0.2 mM
TCEP, and 3 mM R,R⬘-dibromo-m-xylene (Aldrich) in 30% acetonitrile/70% water
mixture] and incubated for 1 h with gentle shaking (22). The cyclized fusions were
eluted, concentrated, and then purified on a Ni-NTA column (Qiagen) under
denaturing conditions (21). The purified fusions were ethanol precipitated and
reverse transcribed (Fig. S1B). In subsequent rounds, translation and fusion formation were performed in 1–2 ml of WG extract and all purification procedures
were done in proportionally smaller volumes.
Selection of IRES RNA Binders. Cyclic peptide-mRNA fusions were applied to an
IRES selection column and incubated for 15–20 min in selection buffer S (20
mM Hepes, pH 7.5, 5 mM MgCl2, 1 mM DTT, 0.05% Tween 20, 10 units/ml
RNasin), supplemented with different concentrations of NaCl, arginine, and
Torula Yeast RNA (TYR) (see below and Table S1). The column was washed
with 10 –20 column volumes of the same buffer and then eluted. The eluted
material was PCR-amplified and used to initiate the next round of selection
(summarized in Table S1).
In rounds 1 and 2 buffer S was supplemented with 0.75 M NaCl, 10 mM
arginine, and 20 ␮M TYR. After washing, mRNA-peptide fusions captured on
the column were eluted by either 10 mM NaOH (round 1) or 8 M urea (round
2). For rounds 3–7, buffer S was supplemented with 0.5 M NaCl, 15 mM
arginine, and 80 ␮M TYR. Preelution was performed in the same buffer with
100 ␮M TYR for 2 h and this eluate was discarded. Specific elution was
performed with 10 ␮M freshly transcribed soluble IRES in buffer S ⫹ 200 mM
NaCl for 2– 4 h at room temperature. In rounds 8 –11, the preelution was
performed in buffer S containing 0.5 M NaCl supplemented with up to 30
␮g/␮l of additional competitor RNA, such as Escherichia coli 16S and 23S RNA
(Roche), phenol-extracted rabbit ribosomes, obtained by gel filtration of RRL
on a Sepharose-6B column (Aldrich), a 319-nt-long mRNA derived from an
unrelated selection, and a Tetrahymena intron RNA in vitro transcribed by T7
RNA polymerase. This eluted material was discarded. The first specific elution
was done with 10 –12 ␮M soluble IRES in buffer S ⫹ 200 mM NaCl for 1–2 h; this
eluate was also discarded. The second specific elution was carried out by using
10 ␮M IRES in buffer S ⫹ 200 mM NaCl for 12–16 h at ⫹4°C and only this eluted
sample was used for the initiation of the next round of selection. PCR products
obtained after rounds 6, 7, and 11 were cloned into the TOPO-TA vector
(Invitrogen) and sequenced. The selection stopped after 11 rounds.
column, and analyzed by MALDI TOF as described (24; see SI Text for details).
Oxidized insulin chain B (Mr 3,495) was used as a mass standard.
Synthetic 6B4 Peptide. The 27-residue peptide MKCSRGIRCAGVLCGSVGHHHHHHHRL (6B4), the 8-residue variant of 6B4 referred to as 6B48, KCSRGIRC, and
the 27-mer 6B4 labeled at its N terminus with the 6-isomer of fluorescein
isothiocyanate (Fl-6B4) were synthesized by using F-moc chemistry and purified by GenScript Corp. Full-length 6B4 was cyclized with dibromo-m-xylene
on a Ni-NTA column as described above, producing a bicyclic derivative of the
peptide, 6B4C. The mass spectra of the peptides were determined by MALDITOF MS (see SI Text for details).
Pepsin Digestion of 6B4C. Linear 6B4 and bicyclic 6B4C peptides were digested
to completion with 0.1% pepsin for 30 min at 30°C in 0.1% TFA at pH ⱕ 2. The
reaction was desalted and concentrated by using C18 Zip Tips (Millipore) and
was analyzed by MALDI-TOF.
Solution Binding Assays. The 35S-labeled peptide 6B4 was synthesized by in
vitro translation, cyclized on a Ni-NTA column when necessary, desalted on a
Sephadex G-10 spin column, and purified on a PepClean C-18 spin-column
(Pierce). For each data point, 200 ␮l of 1 nM linear 6B4 or 0.5 nM cyclic 6B4C
peptide in buffer (20 mM Tris䡠HCl, pH 7.5, 200 mM KCl, 5 mM MgCl2, 0.05%
Triton X-100) was incubated for 1 h with freshly transcribed and purified HCV
IRES RNA. RNA concentration was measured by UV absorption (Cary UV
spectrometer). Equilibrium ultrafiltration measurements of dissociation constants were performed as described in ref. 25.
The fluorescein-labeled peptide Fl-6B4 was used as a probe for solution
binding and competition experiments (see SI Text for details). A sample of 200
␮l of 2 nM Fl-6B4 in buffer (20 mM Hepes, pH 7.4, 300 mM NaCl, 5 mM MgCl2,
2 mM CaCl2, 0.025% Triton X-100, and 0.5% DMSO), was incubated for 1 h with
a series of increasing concentrations of HCV IRES RNA. Equilibrium ultrafiltration was performed by using YM-30 spin filters (Millipore) and fluorescence
spectra of top and bottom chambers were collected on a Varian Cary Eclipse
spectrofluorimeter (see Fig. S3 for examples). The Kd of Fl-6B4 was calculated
as described in ref. 25.
For competitive binding experiments, 200 ␮l of 0.4 – 0.6 nM Fl-6B4 and
15–18 nM IRES (determined to give ⬇70% binding), was preequilibrated for
1 h in binding buffer, then increasing concentrations of competing peptides
were added and incubated for an additional 1 h. Samples were then treated
as described above (see SI Text for details).
IRES-Reporter Gene Constructs. The HCV IRES (1–371) sequence, including 30
nucleotides of the core protein coding region, was added in frame to a
Gaussia luciferase (GLuc, NEB) reporter gene. Control reporter gene constructs were prepared by PCR by using the same GLuc sequences downstream of a 32-mer leader sequence (Fig. 4 A). The constructs were PCR
amplified and used for the transcription of mRNA by T7 RNA polymerase.
Capped constructs were prepared by transcription in the presence of 10
mM cap analog m7GpppG (NEB).
HeLa S10 Extract Preparation and in Vitro Translation in HeLa S10. HeLa S10
translation extract was prepared as described in ref. 14 from a 6-ml HeLa S3 cell
pellet obtained from the National Cell Culture Center. HeLa cell extract
translation reactions were carried out as described in ref. 14. For the measurement of luciferase activity, 50-␮l reactions containing 10 –50 nM reporter
construct mRNAs were incubated for 1 h at 30°C. Different concentrations of
peptides were premixed with measured amounts of mRNA before addition to
in vitro translation extracts. For the visualization of GLuc translation, the
Renilla luciferase assay kit (Promega) was used, because coelenterazine is the
substrate for both Renilla and Gaussia luciferases. Samples of 10 –15 ␮l of
translation reactions were transferred into black 96-well plates (Corning) and
mixed with an equal volume of 1⫻ lysis buffer. The coelenterazine solution in
the assay buffer was added and the light output was measured on a TopCount
NXT luminometer plate reader (Perkin–Elmer).
In Vitro Translation and Cyclization of Peptide 6B4. In vitro translation of the
6B4 peptide was performed in a reconstituted E. coli translation mixture (PURE
system; 23) in the presence of 35S methionine, with cyclization on a Ni-NTA
ACKNOWLEDGMENTS. We thank J. Doudna for the generous gift of HCV IRES
construct; C-W. Lin for the CW mRNA, help with the PURE system, peptide
synthesis, and purification; Z. Mujawar for the pNL4 –3 plasmid; D. Shechner
for the class I ligase RNA; J. Doudna, R. Green, C-W. Lin, D. Treco, Q. Dufton,
M. C. Hartman, F. Seebeck, Y. Guillen, B. Seelig, R. Bruckner, A. Bell, and A.
Luptak for helpful discussions. This work was supported by the HHMI. J.W.S.
is an Investigator and A.L. was an Associate of the Howard Hughes Medical
Institute.
1. Hughes CA, Shafran SD (2006) Chronic hepatitis C virus management: 2000 –2005
update. Ann Pharmacother 40:74 – 82.
2. Simmonds P (2004) Genetic diversity and evolution of hepatitis C virus—15 years on.
J Gen Virol 85:3173–3188.
Litovchick and Szostak
PNAS 兩 October 7, 2008 兩 vol. 105 兩 no. 40 兩 15297
BIOCHEMISTRY
selective enrichment procedures are used. The relatively small
size of 6B48 peptide (Mr of 923 for the linear and 1,025 for the
cyclic form), makes it a promising molecule for further pharmacological optimization.
3. Scott JD, Gretch DR (2007) Molecular Diagnostics of Hepatitis C Virus Infection: A
Systematic Review. JAMA 297:724 –732.
4. Wohnsland A, Hofmann WP, Sarrazin C (2007) Viral determinants of resistance to
treatment in patients with hepatitis C. Clin Microb Rev 20:23–38.
5. Radkowski M, et al. (2005) Persistence of hepatitis C virus in patients successfully
treated for chronic hepatitis C. Hepatology 41:106 –114.
6. Pavio N, Lai MMC (2003) The hepatitis C virus persistence: How to evade the immune
system? J Biosci 28:287–304.
7. Manns MP, et al. (2007) The way forward in HCV treatment—Finding the right path.
Nat Rev Drug Discov 6:991–1000.
8. Pestova TV, et al. (1998) A prokaryotic-like mode of cytoplasmic eukaryotic ribosome
binding to the initiation codon during internal translation initiation of hepatitis C and
classical swine fever virus RNAs. Genes Dev 12:67– 83.
9. Hellen CUT, Sarnow P (2001) Internal ribosome entry sites in eukaryotic mRNA molecules. Genes Dev 15:1593–1612.
10. Martınez-Salas E, Pacheco A, Serrano P, Fernandez N (2008) New insights into internal
ribosome entry site elements relevant for viral gene expression. J Gen Virol 89:611–
626.
11. Dash S, et al. (2005) Interferons ␣, ␤, ␥ each inhibit hepatitis C virus replication at the
level of internal ribosome entry site-mediated translation. Liver Int 25:580 –594.
12. Fraser CS, Doudna JA (2007) Structural and mechanistic insights into hepatitis C viral
translation initiation. Nat Rev Microbiol 5:29 –38.
13. Spahn CMT, et al. (2001) Hepatitis C virus IRES RNA–induced changes in the conformation of the 40S ribosomal subunit. Science 291:1959 –1962.
14. Otto GA, Puglisi JD (2004) The pathway of HCV IRES-mediated translation initiation.
Cell 119:369 –380.
15. Gallego J, Varani G (2002) The hepatitis C virus internal ribosome-entry site: A new
target for antiviral research. Biochem Soc Trans 30(Part 2):140 –145.
16. Seth PP, et al. (2005) SAR by MS: Discovery of a new class of RNA-binding small
molecules for the hepatitis C virus: Internal ribosome entry site IIA subdomain. J Med
Chem 48:7099 –7102.
17. Jefferson EA, et al. (2004) Biaryl guanidine inhibitors of in vitro HCV-IRES activity.
Bioorg Med Chem Lett 14:5139 –5143.
15298 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0805837105
18. Kim JJ (2007) Antiviral compositions and methods of using the same. US patent
application 20070259844.
19. VGX Pharmaceuticals, Inc. (2005–2007). A randomized, open-label phase II trial of the
anti-HCV activity and safety of VGX-410 (Mifepristone) at 3 dose levels in HCV infected
patients. Phase II Clinical Trials NCT00255177.
20. Roberts RW, Szostak JW (1997) RNA-peptide fusions for the in vitro selection of
peptides and proteins. Proc Natl Acad Sci USA 94:12297–12302.
21. Liu R, Barrick JE, Szostak JW, Roberts RW (2000) Optimized synthesis of RNA-protein
fusions for in vitro protein selection. Methods Enzymol 318:268 –293.
22. Timmerman P, Beld J, Puijk WC, Meloen RH (2005) Rapid and quantitative cyclization
of multiple peptide loops onto synthetic scaffolds for structural mimicry of protein
surfaces. ChemBiochem 6:821– 824.
23. Shimizu Y, et al. (2001) Cell-free translation reconstituted with purified components.
Nat Biotechnol 19:751–755.
24. Seebeck FP, Szostak JW (2006) Ribosomal synthesis of dehydroalanine-containing
peptides. J Am Chem Soc 128:7150 –7151.
25. Davis JH, Szostak JW (2002) Isolation of high-affinity GTP aptamers from partially
structured RNA libraries. Proc Natl Acad Sci USA 99:11616 –11621.
26. Litovchick A, Rando RR (2003) Stereospecificity of short Rev-derived peptide interactions with RRE IIB RNA. RNA 9:937–948.
27. Barrick JE, et al. (2001) Large libraries reveal diverse solutions to an RNA recognition
problem. Proc Natl Acad Sci USA 98:12374 –12378.
28. Rogers GA, Shaltiel N, Boyer PD (1976) Facile alkylation of methionine by benzyl
bromide and demonstration of fumarase inactivation accompanied by alkylation of a
methionine residue. J Biol Chem 251:5711–5717.
29. Athanassiou Z, et al. (2007) Structure-guided peptidomimetic design leads to nanomolar beta-hairpin inhibitors of the Tat-TAR interaction of bovine immunodeficiency
virus. Biochemistry 46:741–751.
30. Leeper TC, et al. (2005) TAR RNA recognition by a cyclic peptidomimetic of Tat protein.
Biochemistry 44:12362–12372.
31. Bartel DP, Szostak JW (1993) Isolation of new ribozymes from a large pool of random
sequences. Science 261:1411–1418.
Litovchick and Szostak